In a service, client-initiated requests often result in the generation of tasks or various workloads for execution by computing systems or other computing resources that effectuate the service. In addition, satisfaction of the requests may be contingent on the completion of the tasks or the workloads. Furthermore, the timeliness of the completion of a task may change due at least in part to environmental conditions, such as the availability of computing resources to take up task execution, the computational complexity associated with the task and the number of tasks competing for the computing resources, among others. An undesirable effect of the dependence on the environmental conditions may be manifested in clients observing variability in the amount of time required for their requests to be satisfied.
It is often challenging to configure a computing system to provide predictability in request completion. Furthermore, it is also challenging to provide for fairness among the various clients of the computing system.